GPS (Global Positioning System) is an earth-satellite-based electronic system for enabling GPS receivers in ships, aircraft, land vehicles and land stations to determine their geographic and spatial position such as in latitude, longitude, and altitude. Discussion of GPS herein is without limitation to other analogous electronic systems as well as applicable receiver circuits in a variety of telecommunication systems. “GNSS” herein refers to any navigation satellite system. A GNSS receiver computes the user position by triangulation or trilateration using measured distances to enough satellite vehicles (earth satellites) SVi to achieve a position fix. “Assisted GNSS” assists the computations by information obtained from a terrestrial communications network.
Glonass support to receive Glonass (Russian) satellites is also fast becoming an additional key requirement for GPS receivers in the USA and much of the rest of the world. In FIG. 1, users care about sensitivity performance of satellite positioning receivers because such sensitivity enables the receiver to work indoors more often and improves the user experience outdoors and indoors. For receiving GNSS and assisted GNSS, every 0.5 dB of receiver sensitivity is valuable, e.g. desirably enhancing performance margin for industry tests like those of the group 3GPP.
It would be desirable to even more accurately, reliably, rapidly, conveniently and economically search for, acquire, and track received signals and maintain accurate time, position, velocity, and/or acceleration estimation in a communication device having a satellite positioning receiver (SPR) or other receiver and its clock source.
In FIG. 2 for a typical satellite acquisition strategy, a terrestrial receiver 100 has an RF front end 110, and baseband signal processing BSP/Correlators 120 section accumulates correlation results across several milliseconds for satellite signal detection. Accumulation of correlation results is done using a combination of coherent integration (a.k.a ‘pre-detection’ PreD) and non-coherent integration (a.k.a ‘post-detection’ PostD) in an accumulate and dump (AD) section 130 followed by final processing and navigation applications 140. Such a process of correlation and accumulation is called a “dwell” and has a dwell time TD where TD=PreD*PostD.
Coherent integration boosts signal-to-noise ratio SNR by adding (accumulating or integrating) the 1 ms repetitions of the satellite signal coherently. The signal is e.g. binary BPSK (binary phase shift keying) modulated with +/−1 modulation, meaning +/−180 degrees phase shift depending on each bit. Coherent accumulation adds the received signal waveform of each repetition (e.g., each 1 ms time window) arithmetically over a number of repetitions within a given bit interval, while noise being statistical accumulates in rms (root-mean-square) value more slowly as the square root of the number of repetitions so that SNR is boosted.
Non-coherent integration squares the signal (S2=S(t)×S(t)) to remove the BPSK modulation (i.e., the square of +/−1 modulation is always +1). This way, non-coherent integration can be performed over many bit intervals to increase the overall gain. However, non-coherent integration also squares the noise N(t) along with the signal S(t). Heuristically, consider a summation over time of squaring combined signal and noise (S(t)+N(t))2=S2(t)+2S(t)N(t)+N2(t). The summing basically builds up S2(t)+N2(t) more than the cross-product term. So SNR is generally not increased as much by the non-coherent integration as by coherent integration. Strictly speaking, even after squaring, the squared-signal S2(t) linearly increases with integration whereas standard deviation of squared-noise N2(t) increases more slowly. The squaring removes phase information, which leads to loss of SNR. The relatively-poorer noise performance of non-coherent integration compared to coherent integration is called squaring loss. (Analogous remarks directed specifically to an absolute value (abs) approach to noncoherent integration could also be stated here but are omitted, because like squaring, abs operates on only same-signed sample values.)
For a given dwell time TD, longer coherent integration time PreD means more SNR enhancement and fewer terms (i.e. equal in number to PostD ratio of TD/PreD) in the non-coherent integration or accumulation. For example, a receiver can use a dwell with PreD=1 ms and PostD=200 to detect signals down to −140 dBm. Receivers can use a dwell with PreD=20 ms and PostD=800 to detect signals down to or as low as −160 dBm.
Longer coherent integration time (PreD, e.g. 20 ms) reduces squaring loss associated with non-coherent operation and hence achieves better receiver sensitivity.
GPS and other GNSS acquisition sensitivity is limited by the longest coherent integration period (PreD) possible for a given integration time. In GPS signal structure, each bit duration is 20 ms. See FIG. 3A, and magnified details in FIGS. 3B and 3C.
High-sensitivity dwells are usually done using the longest coherent integration period (PreD) possible. However, the bit-edge alignment is unknown at the beginning. See FIG. 3A. Therefore, whenever the coherent integration straddles a bit boundary, sensitivity is likely to suffer from bit-edge transition-related loss. This is because coherent accumulation of the waveform may result in some waveform subtraction near such a bit boundary if the BPSK modulation transitions from +1 to −1, or from −1 to +1, there. In GPS, one possible technique is to use a PreD of 19 ms when bit edge is not known. The number 19 is co-prime with the number 20 (i.e., 19 and 20 are relatively prime), which thereby positions a 19 ms window repeatedly across a long dwell so that the 19 ms window variously straddles 20 ms bit widths or occasionally lies in one of them. Therefore the loss due to integrations across bit-edge transitions over a long dwell approaches an average loss value, which can be computed or estimated predictively from the respective losses of each of many positions of the 19 ms window across the dwell.
For GPS, the expected or average de-sense due to PreD of 19 ms is around 1.6 dB as compared to an ideal bit-aligned PreD of 20 ms. (De-sense refers to a number of dB of diminished sensitivity relative to the sensitivity which would be enjoyed under an ideal receiver processing condition or some receiver processing condition used as a reference.) A first problem accordingly confronts the art, namely how can the GPS or other GNSS acquisition sensitivity be improved further when the bit boundaries are not known?
Another problem is that even if bit edge is known, it can be desirable to restrict the duration of coherent integration period PreD to reduce the number of Doppler searches or have more protection against clock or user dynamics. But sensitivity is limited by restricting PreD.
For Glonass, the same conventional approach would use PreD of 9 ms when bit edge is not known since the Manchester code (FIG. 8) used in Glonass causes a periodic flip (i.e., a further multiplication by minus-one) every 10 ms. Average (expected) de-sense for Glonass because of unknown bit edge for PreD of 9 ms is an undesirably-large˜4 dB de-sense as compared to ideal performance with a bit-aligned PreD of 20 ms. A corresponding problem thus also confronts the art, namely how can the Glonass acquisition sensitivity be improved when the bit boundaries are not known and given the periodic flip?
Accordingly, substantial departures in GNSS receiver and other spread spectrum receiver technology are sought and would be most desirable and beneficial in this art.